Using LowSpeed Links for HighSpeed Wireless Data Delivery

1
Using Low-Speed Links for High-Speed Wireless
Data Delivery

• Henning Schulzrinne
• Dept. of Computer Science
• Columbia University
• (with Stelios Sidiroglou and Maria Papadopouli)

2
Overview

• Disconnected ad-hoc networks
• multi-modal networking
• using low-speed feedback to accelerate data
  delivery
• 7DS prototype
• future work

3
Wireless Network filling the infrastructure-ad
hoc gap

• Wireless networks
• Ubiquitous, fast, cheap pick any two
• Currently, varies from 0.1c to 4/MB
• Research has primarily explored
• one-hop infrastructure extension (2G, 3G, 802.11)
• multi-hop connected ad-hoc networks (mesh
  networks)
• But
• 2G/3G bandwidth will remain low and precious
• hot spots not ubiquitous
• ad hoc networks dont scale
• brittle if spanning large areas
• Our proposal use mobile nodes to carry data
• to and from infrastructure networks

4
Cost of networking
5
Limitations of 802.11

• Good for hotspots, difficult for complete
  coverage
• Manhattan 60 km2 ? 6,000 base stations (not
  counting vertical)
• With 600,000 Manhattan households, 1 of
  households would have to install access points
• Almost no coverage outside of large coastal cities

6
7DS a framework for intermittently connected
networks

• Two directions for data
• Internet ? mobile nodes
• mobile nodes ? Internet
• Each in multiple hops
• but not routed

delay
bandwidth (peak)
7DS seven degrees of separation
7
Applications

• Tourism
• get information about sights, travel, public
  transport schedules, ..
• upload picture postcards and video recordings
• Transportation
• users in buses and trains leverage data
  capability
• Emergencies
• propagate Im alive and rescue information
• Mobile sensors
• sensors spread too far to communicate directly
  with each other
• large sensor data objects

8
A family of access points
9
Network to Mobile

• Deliver web content to roaming user

weather?
web cache
query for all documents
multicast
7DS node
deliver matching documents
10
Simulation environment
pause time 50 s mobile user speed 0 .. 1.5
m/s host density 5 .. 25 hosts/km2 wireless
coverage 230 m (H), 115 m (M), 57.5 m
(L) ns-2 with CMU mobility, wireless
extension randway model
querier
wireless coverage
dataholder
randway model
11
Average Delay (s) vs Dataholders ()Peer-to-Peer
schemes
high transmission power
medium transmission power
12
Modeling

• Carrier is infected, hosts are susceptible
• Transmit to any give host with probability
  hao(h) in interval h
• Pure birth process
• Ttime until data has spread among all mobiles
• ET1/a S

• Statistical mechanics model can accurately
  predict data distribution for some scenarios

N-1
i1
13
Mobile to Internet

• Email service interface

propagate to other pedestrians
7DS MTA
encrypt message encrypt headers with 7DS public
key
14
Realization
15
Closing the loop in 7DS

• Problems with open-loop propagation systems
• Network to mobile
• no way to inject popular content into the system
• Mobile to network
• have to limit replication to avoid flooding
• If too few copies, may never get delivered
• copies persist long after delivery succeeded

• Thus, transform into closed-loop system
• dont know who needs information
• but likely regionally limited by mobility
• ? regional broadcast of control information
• no need for bidirectional data
• low bandwidth

16
Options for closing the loop

• Options
• satellite radio (XM, Sirius)
• LEO satellites (Iridium)
• low-bandwidth cellular (CDPD, GSM)
• one-way or two-way pagers

See also Ambient Devices
17
Pagers as feedback channel
FLEX 1600-6400 b/s
remove from cache
SNPP (RFC 1861)
MTA
message 42 delivered
PL-900 POCSAG
18
Cache management details

• Receiving MTA broadcasts unique (hash) identifier
  of message
• hash long enough to prevent spoofing
• 7DS nodes remove from cache
• other MTAs prevent delivery
• Popularity management
• indications of popular content distributed to 7DS
  nodes
• nodes query that content from others

• Reputation management
• distribute identifier for good and bad guys
• good guys deliver messages fast
• bad guys never deliver messages
• accept messages preferably from good guys

19
Current status prototype

• Initial Java implementation
• search not just by URL, but by content
• ? greater likelihood of finding appropriate
  material (news)
• Working on PDA implementations
• Also, considering Linux embedded systems
• low-power, self-contained

20
7DS node
21
On-going work leveraging low-bandwidth links

• Hordes of low-bandwidth nodes
• split large or urgent message into pieces
• spread pieces across many nodes
• each node transmits at very low rate
• use Tornado codes for redundancy
• cf. BitTorrent

22
Conclusion

• 7DS as extension of infrastructure and ad-hoc
  networks
• Combine benefits of low bit-rate, but ubiquitous
  and high bit-rate, but sparse networks